cleaner graph nodes

mcp_codebase.py

@@ -2101,11 +2101,24 @@ def build_indexes():
         if not name or node_type == "unknown" or not fpath:
             continue
 
-        # Ensure file node exists
+        # ---- File node ---------------------------------------------
-        file_node_id = f"file::{fpath}"
+        # Keep the same ID (`file::<path>`) but store the full set of
+        # attributes so that downstream code can rely on `name`, `lang`,
+        # and `file` (even though `name` will just be the file path).
         if file_node_id not in graph.graph.nodes:
-            graph.add_node(file_node_id, type="File", path=fpath, lang=lang)
+            graph.add_node(
-            created_nodes[file_node_id] = {"type": "File", "path": fpath}
+                file_node_id,
+                type="File",
+                name=fpath,
+                path=fpath,
+                lang=lang,
+            )
+            created_nodes[file_node_id] = {
+                "type": "File",
+                "name": fpath,
+                "path": fpath,
+                "lang": lang,
+            }
 
         # Create node ID with language
         node_id = f"{lang}::{node_type}::{fpath}::{name}"
@@ -2260,8 +2273,29 @@ def build_indexes():
             if not candidate_id:
                 # If nothing exists, fall back to a generic import node.
                 candidate_id = f"import::{imp}"
+                # ---- Import node --------------------------------------------
+                # For imports we want a node that mirrors the target module if
+                # it exists; otherwise we still create a stub with language &
+                # file attributes for consistency.
                 if candidate_id not in graph.graph.nodes:
-                    graph.add_node(candidate_id, type="Import", name=imp)
+                    # First try to find a module node with that name
+                    found = False
+                    for n_id, n_data in created_nodes.items():
+                        if n_data.get("name") == imp and n_data.get("type") == "module":
+                            candidate_id = n_id
+                            found = True
+                            break
+                    if not found:
+                        # Generic import stub
+                        candidate_id = f"import::{imp}"
+                        graph.add_node(
+                            candidate_id,
+                            type="Import",
+                            name=imp,
+                            lang="unknown",
+                            file="",
+                        )
+                graph.add_edge(file_node_id, candidate_id, "imports")
 
             graph.add_edge(file_node_id, candidate_id, "imports")
 
@@ -2277,8 +2311,32 @@ def build_indexes():
             if not target_id:
                 # Fallback to a generic call node.
                 target_id = f"call::{call}"
+                # ---- Call node ---------------------------------------------
+                # If we already know the target node (by name), reuse it.  Otherwise
+                # create a generic node that still carries language & file context
+                # so it can be displayed or filtered later.
                 if target_id not in graph.graph.nodes:
-                    graph.add_node(target_id, type="FunctionCall", name=call)
+                    # Attempt a best‑effort lookup for the actual function node
+                    found = False
+                    for n_id, n_data in created_nodes.items():
+                        if (
+                            n_data.get("name") == call
+                            and n_data.get("type") in {"function", "method", "constructor"}
+                        ):
+                            target_id = n_id
+                            found = True
+                            break
+                    if not found:
+                        # Fallback to a generic FunctionCall node
+                        target_id = f"call::{call}"
+                        graph.add_node(
+                            target_id,
+                            type="FunctionCall",
+                            name=call,
+                            lang="unknown",
+                            file="",
+                        )
+                graph.add_edge(node_id, target_id, "calls")
 
             graph.add_edge(node_id, target_id, "calls")